Re: [Xen-devel] Xen 4.6 Development Update (two months reminder)

[Meng Xu <xumengpanda@xxxxxxxxx>]

2015-03-12 11:39 GMT-04:00 Dario Faggioli <dario.faggioli@xxxxxxxxxx>:
>
> On Thu, 2015-03-12 at 15:07 +0000, Ian Campbell wrote:
> > On Thu, 2015-03-12 at 10:21 +0000, wei.liu2@xxxxxxxxxx wrote:
> >
> > > *  Repurpose SEDF Scheduler for Real-time (fair)
> > >    RFC patch posted (v2)
> > >   -  Joshua Whitehead, Robert VanVossen
> >
> > This was superceded by the RTDS stuff, wasn't it?
> >
> I haven't head from Joshua and Robert in a while, so I don't really know
> whether they're still working on this, but I think they're not.
>
> And yes, we have a new and modern real-time scheduling now, so I would
> rather direct all the effort toward it (to move it out from
> 'experimental' status), and start thinking at ways to deprecate/get rid
> of SEDF.


Dagaen, Chong and I are working on improving the RTDS scheduler right now.

Dagaen will change it from quantum driven to timer based scheduler;

Chong will extend the xl toolstack to support per-vcpu setting/getting
based on our old patch set.

I'm working on eliminating the scheduler overhead for the dedicated
VCPU (A VCPU is dedicated VCPU if it has full capacity, and its hard
affinity has only one CPU and user choose to set it as dedicated.).
Once a VCPU is dedicated to a CPU, no other VCPUs will run on that
CPU. So there is no point to trigger the scheduler on that CPU, which
cause around 1500cycles schedule() overhead. I'm working on the code
and then evaluating its performance. Once I can confirm this will
definitely benefit the RTDS scheduler, I will send a separate email to
the ML to explain the goal/design/implementation in detail.

Thanks,

Meng


-- 


-----------
Meng Xu
PhD Student in Computer and Information Science
University of Pennsylvania

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel